230 METALLURGY OP IKON AND STEEL. series is calculated to give the same weight per ton of pig-iron as for the first series. In the discussion of Mr. Talbofs paper, Mr. Monell gave figures of the work at Homestead, but the data were not complete and a calculation along the same lines as the foregoing leaves 5.4 per cent, of metallic iron unaccounted for. Mr. Eartshorne* gives a summary for the work at Kladno, but this also is incomplete and the figures indicate that 8.2 per cent, has disappeared. It is only by the most careful weighing that the records can be'of value on this question of loss, for it is easy to make a mistake of one per cent, in weighing the stock or the ingots. The difference between a gain of 3 per cent, and 4= per cent, in an open-hearth furnace is a very important matter, but it is necessary to find out whether it is in the operation of the furnace or in keeping the accounts. When the -loss is found by subtracting the product from the stock used, it is as if we should determine the percentage of silicon in pig-iron by determining the phosphorus, manganese, sulphur, copper and metallic iron, and then subtracting their sum from one hundred and calling the remainder silicon. Every one recognizes the error involved in a "determination by difference." This method has its uses, and the determination is correct within certain limits, but it must not be accepted too implicitly. In important investigations the slag should be weighed and analyzed, and if the loss of metallic iron in the slag agrees with the iron not otherwise accounted for, there is a check on the whole calculation showing that the weights are right for both metal and slag. The results given by Mr. Talbot answer these conditions and are quoted here as corroborative of the experiments made at Steelton. The whole matter of gain and loss in open-hearth practice is a question of terms. Usually the weight of the ore is not reckoned. Thus in a heat of all pig-iron, there will be 50 tons of iron and 12 tons of ore, and if the ingots weigh 50 tons we say the loss is nil, disregarding the-12 tons of ore containing 7 tons of metallic iron. If, on the other hand, we add the weight of the ore, we are again wrong, for this ore contains 5 tons of oxygen, silica and water. If the actual content of metallic iron be calculated in the ore addition, then the percentage of water must be allowed for, and if this refinement be carried out, then we must subtract the carbon and * Trans, A. I. If. E., February, 1900.